Aberrant sperm DNA methylation predicts male fertility status and embryo quality

ORIGINAL ARTICLE: ANDROLOGY 1 2 3 4 5 6 7 8 9 10 a b a c Q20 Kenneth I. Aston, Ph.D., Philip J. Uren, Ph.D., Timothy G. Jenkins, Ph.D., Alan Horsager, Ph.D., 11 Bradley R. Cairns, Ph.D.,d,e Andrew D. Smith, Ph.D.,b and Douglas T. Carrell, Ph.D.a,f 12 a Department of Surgery, University of Utah Andrology and IVF Laboratories, University of Utah School of Medicine, Salt 13 Lake City, Utah; b Molecular and Computational Biology, University of Southern California, Los Angeles, California; c 14 Episona, Inc, Glendale, California; d Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah 15 School of Medicine, Salt Lake City, Utah; e Howard Hughes Medical Institute, Chevy Chase, Maryland; and f Department of Obstetrics and Gynecology and Department of Human Genetics, University of Utah School of Medicine, Salt Lake City, 16 Utah 17 18 19 20 Objective: To evaluate whether male fertility status and/or embryo quality during in vitro fertilization (IVF) therapy can be predicted 21 based on genomewide sperm deoxyribonucleic acid (DNA) methylation patterns. 22 Design: Retrospective cohort study. 23 Setting: University-based fertility center. 24 Patient(s): Participants were 127 men undergoing IVF treatment (where any major female factor cause of infertility had been ruled 25 out), and 54 normozoospermic, fertile men. The IVF patients were stratified into 2 groups: patients who had generally good embryo26 genesis and a positive pregnancy (n ¼ 55), and patients with generally poor embryogenesis (n ¼ 72; 42 positive and 30 negative preg27 nancies) after IVF. Intervention(s): Genomewide sperm DNA methylation analysis was performed to measure methylation at >485,000 sites across the 28 genome. 29 Main Outcome Measure(s): A comparison was made of DNA methylation patterns of IVF patients vs. normozoospermic, fertile men. 30 Result(s): Predictive models proved to be highly accurate in classifying male fertility status (fertile or infertile), with 82% sensitivity, 31 and 99% positive predictive value. Hierarchic clustering identified clusters enriched for IVF patient samples and for poor-quality–em32 bryo samples. Models built to identify samples within these groups, from neat samples, achieved positive predictive value R94% while 33 identifying >one fifth of all IVF patient and poor-quality–embryo samples in each case. Using density gradient prepared samples, the same approach recovered 46% of poor-quality–embryo samples with no false positives. 34 Conclusion(s): Sperm DNA methylation patterns differ significantly and consistently for infertile vs. fertile, normozoospermic men. In 35 addition, DNA methylation patterns may be predictive of embryo quality during IVF. (Fertil 36 Use your smartphone SterilÒ 2015;-:-–-. Ó2015 by American Society for Reproductive Medicine.) 37 to scan this QR code Key Words: Sperm DNA, DNA methylation, IVF outcome, embryo, genomewide, microarray, 38 and connect to the male infertility discussion forum for 39 this article now.* Discuss: You can discuss this article with its authors and with other ASRM members at http:// 40 fertstertforum.com/astonk-sperm-dna-methylation-infertility/ 41 * Download a free QR code scanner by searching for “QR scanner” in your smartphone’s app store or app marketplace. 42 43 44 45 he mainstay of male infertility grading, semen analysis has changed parameters evaluated by the standard 46 diagnosis is the standard semen very little over the past several decades. analysis (1–3). Except for severely 47 analysis. With the exception of Numerous studies have evaluated the diminished sperm count or motility, 48 modification of criteria for morphologic prognostic value of the various semen the predictive value of semen analysis 49 for fertility is modest at best. A 50 Q2 milestone study of the predictive value 51 Received June 1, 2015; revised July 29, 2015; accepted August 18, 2015. of semen analysis concluded that K.I.A., P.J.U., and T.G.J. have a patent in preparation entitled Methods of Identifying Male Fertility 52 Status and Embryo Quality. A.H. is licensing this technology with Episona, Inc. B.R.C. has a patent although it is useful for classifying 53 in preparation entitled Methods of Identifying Male Fertility Status and Embryo Quality. A.D.S. men as being either subfertile, of and D.T.C. are shareholders in Episona, Inc. 54 Q1 Reprint requests: Douglas T. Carrell, Ph.D., Division of Urology-Andrology/IVF Laboratories, University indeterminate fertility, or fertile, it is 55 of Utah School of Medicine, Andrology, 675 S Arapeen Dr, Ste #205, Salt Lake City, Utah 84108 very ineffective for diagnosing 56 (E-mail: [email protected]). infertility, owing to the fact that semen 57 Fertility and Sterility® Vol. -, No. -, - 2015 0015-0282/$36.00 parameters for many infertile men fall 58 Copyright ©2015 American Society for Reproductive Medicine, Published by Elsevier Inc. within normal ranges (4). http://dx.doi.org/10.1016/j.fertnstert.2015.08.019 59

Aberrant sperm DNA methylation predicts male fertility status and embryo quality

T

VOL. - NO. - / - 2015

1

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118

ORIGINAL ARTICLE: ANDROLOGY 119 The main parameters evaluated in semen analysis, 120 namely sperm count, motility, viability, and morphology, 121 are somewhat subjective; consequently, they are subject to 122 technical error. Although continual training and assessment, 123 quality control measures, and proficiency testing all minimize 124 technical error, multiple studies have demonstrated that coef125 ficients of variation (CVs) between labs and technicians 126 commonly fall in the 20%–30% range, with higher CVs re127 ported in some studies (5–7). In addition to the technical 128 variability inherent in the testing, semen parameters for the 129 same individual vary significantly among collections, with 130 CVs of approximately 30% between any 2 collections from 131 Q3 the same man, according to a recent study of >5,000 men 132 (8). Given this inherent variability, the World Health 133 Organization recommends that R2 semen analyses be 134 performed before clinical decisions are made (9). 135 Lastly, the predictive value of the various semen param136 Q4 eters has been demonstrated to be severely limited. Two large 137 and comprehensive studies have been performed to charac138 terize semen parameters in healthy, fertile men; fertile and 139 subfertile ranges have been defined for each parameter as140 sessed (4, 10). Nevertheless, assessment using standard 141 semen analysis is broadly accepted to fall far short of the 142 goal of predicting fertility potential. 143 Adjunct tests have been developed over the years, such as 144 sperm deoxyribonucleic acid (DNA) damage assessment, 145 capacitation, and acrosome reaction tests, egg and zona pene146 tration assays, antisperm antibody testing, aneuploidy 147 screening, motile sperm organelle morphology examination, 148 and hyaluronic acid binding (11–18). Although these tests 149 can be helpful in characterizing fertility potential in select 150 patients, the predictive values of the assays are generally 151 accepted as being suboptimal (19). The need for additional 152 diagnostic tools for evaluation of male infertility is widely 153 acknowledged (2). 154 With advancing molecular diagnostic tools, the identi155 fication of novel genetic and epigenetic markers of male 156 infertility is becoming a realistic option. A few genetic 157 markers for male infertility have been identified, such as 158 Y-chromosome microdeletions (20), Klinefelter syndrome 159 (21), and DPY19L and SPATA16 mutations (22–25), 160 among others. These genetic features are associated 161 with extreme male infertility phenotypes, including 162 severe oligozoospermia, nonobstructive azoospermia, and 163 complete globozoospermia, accounting for a very small 164 percentage of infertile men. 165 Although genetics refers to the DNA sequence itself, epi166 genetics refers to modifiable but generally stable modifica167 tions to the DNA or chromatin packaging. Direct DNA 168 modifications consist of methylation alteration to the 5169 carbon position of cytosine bases, usually in the context of 170 cytosine-guanine dinucleotides (CpGs). Chromatin packaging 171 modifications include covalent modifications to histones, and 172 in sperm, the distribution of histones and protamines in the 173 genome. Epigenomic information can be reliably replicated 174 through cell divisions, and although the epigenome is largely 175 reprogrammed in early gametogenesis and embryogenesis, 176 some portion of the epigenome in parental gametes is trans177 mitted to offspring. The amount of epigenomic information

that is heritable in humans is largely unknown, as are the functions and consequences of epigenomic inheritance. Sperm epigenetics is an emerging area of study, driven largely by early observations that the way DNA is packaged within a sperm head may affect the capacity of sperm to fertilize an egg and/or the developmental capacity of the Q5 zygote (26–29). Sperm DNA has long been recognized as being packaged differently, compared with other cell types (namely by protamines). Protamine packaging was generally assumed to serve utilitarian purposes, including compaction of the sperm nucleus to facilitate efficient motility, and protection of sperm DNA from the harsh environment of the female reproductive tract (30). However, more recent characterization of the epigenetic landscape of the sperm nucleus, specifically the localization of, and modification to, histones that remain associated with sperm DNA after replacement of most histones by protamines, indicates that sperm epigenetic architecture reflects the development of mature sperm from spermatogonial stem cells and likely contributes to early embryonic development (31, 32). Numerous recent studies have reported abnormal sperm DNA methylation patterns associated with infertility, particularly in oligozoospermic patients. However, most of these studies evaluated just 1 or a few genes, or included very small cohorts using genomewide approaches (33–48). The aim of the current study was to evaluate genomewide sperm DNA methylation patterns in a large cohort of sperm donors and in vitro fertilization (IVF) patients, to determine whether DNA methylation patterns are predictive of fertility status or IVF prognosis. The current study was motivated by 3 primary factors. (1) The current diagnostic standard for male infertility (i.e., semen analysis) provides minimal clinically actionable information. (2) Numerous exploratory studies have established Q6 that altered sperm DNA methylation in single or multiple genes is associated with male infertility. (3) Sperm DNA methylation patterns are relatively stable within an individual (49). The goal of the current study was to determine whether sperm DNA methylation patterns can be used to predict male fertility status and IVF success.

MATERIALS AND METHODS Sample Selection Semen samples used for the current study were obtained from the University of Utah tissue bank, after informed consent had been provided according to IRB-approved protocols. Individuals were asked to adhere to general semen collection instructions, which included 2–5 days of abstinence immediately preceding collection. Collected samples were mixed in a 1:1 Q7 ratio with Test Yolk Buffer (Irvine Scientific) and stored in liquid nitrogen until they were used in this study.

Controls Control samples (n ¼ 54) were collected from normozoospermic men with proven fertility. Most control samples were composed of whole ejaculate (n ¼ 36), whereas 12 were VOL. - NO. - / - 2015

2

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236

Fertility and Sterility® 237 prepared by density-gradient centrifugation prior to cryo238 preservation. For 6 samples, the preparation method was 239 not recorded. 240 241 Patient Samples 242 Samples were selected from 292 IVF patients based on embryo 243 quality and pregnancy outcome (Supplemental Table 1, avail244 able online). Couples with moderate-to-severe female factor 245 infertility, including advanced maternal age and severe endo246 metriosis or polycystic ovarian syndrome, were excluded 247 from the study. A total of 55 patients were selected because 248 they had high embryo quality overall, and a confirmed chem249 ical pregnancy. Embryos were scored based on previously 250 reported criteria relating to blastomere number and fragmen251 tation in early embryos, and trophectoderm and inner cell252 mass quality in blastocysts (50). 253 Fetal heartbeat was detected in 49 patients (89.1%), unde254 tectable in 4 patients (7.3%), and not recorded in 2 patients 255 (3.6%). Seventy-two patients were selected who displayed 256 generally poor embryogenesis, including increased rates of 257 early or late developmental arrest, or reduced embryo quality 258 due to blastomere fragmentation. Of these, 42 achieved a 259 pregnancy; 30 did not. Of the 42 pregnant patients, fetal 260 heartbeat was detected in 33 (78.6%), undetectable in 8 261 (19%), and not recorded in 1 patient (2.4%). Live-birth 262 outcome data were not available for all patients, but for pa263 tients for whom data was recorded, 75% in the ‘‘good embryo’’ 264 group delivered successfully, compared with only 31.4% in 265 the ‘‘poor embryo’’ group. 266 267 268 Gradient-Prepared Patient Samples 269 A separate experiment was performed on a subset of the sam270 ples from IVF patients. For samples that contained >5  106 271 progressively motile sperm (n ¼ 44), DNA methylation was 272 assessed on the whole ejaculate; separately, a portion of the 273 sample was purified using a 45%/90% discontinuous isolate 274 gradient, and the 90% fraction was subjected to DNA methyl275 ation analysis as well. 276 277 278 Patient Details 279 Supplemental Table 2 (available online) presents the fre280 quency of male factor infertility (defined as being below the 281 World Health Organization threshold in R1 of the semen pa282 rameters), female factor infertility, the presence of both male 283 and female factors, and the designation of idiopathic infer284 tility among both partners. Shown in addition is the nature 285 of the various mild female factors within each group of IVF 286 patients. The frequencies of all factors are statistically similar 287 among groups (P>.05; c2 analysis). 288 289 Sample Preparation 290 291 Q8 For DNA isolation, sperm samples were thawed simulta292 neously and were subjected to a column-based DNA extrac293 tion protocol with sperm-specific modification to the 294 DNeasy kit (Qiagen). Prior to DNA extraction, somatic cell 295 lysis was performed by incubation in somatic cell lysis buffer

(0.1% sodium dodecyl sulfate, 0.5% Triton X-100 in diethylpyrocarbonate H2O) for 20 minutes on ice, to eliminate white blood cell contamination. After somatic cell lysis, the sperm were pelleted, and a visual inspection of each sample was performed to ensure the absence of all potentially contaminating cells before proceeding. For bisulfite conversion and array processing, extracted sperm DNA was bisulfite converted with the EZ-96 DNA Methylation-Gold kit (Zymo Research), according to manufacturer recommendations specifically for use with array platforms. The converted DNA was delivered to the University of Utah Genomics Core Facility and hybridized to Infinium HumanMethylation450 BeadChip microarrays (Illumina) and analyzed according to manufacturer protocols.

Statistical Analysis After the hybridization protocol, arrays were scanned. The Q9 minfi software package (Bioconductor; (51)) was used to generate b-values (so called because they are expected to follow a b-distribution, each is a value between zero and 1, for each probed CpG, representing the proportion of DNA molecules that are methylated at the given locus). The same package was used to apply subset-quantile within array normalization (52) to the extracted b-values. Statistical comparisons were made as follows: (1) all patients vs. controls; and (2) IVF patients with good vs. poor embryogenesis. Hierarchic clustering was applied using the Euclidean distance between methylation profiles. For construction of discriminative models to differentiate IVF patient samples from fertile donor samples, and good from poor-quality–embryo samples, we evaluated 2 types of features to describe each sample: the Euclidean distance between the sample and all other samples in the training set, and a subset of the individual CpG methylation values for each sample. In the latter case, to select the subset, we first identified the 500 most-discriminative loci using Wilcoxon’s rank sum test (i.e., the 500 loci with the lowest P values). All models were constructed (including feature-selection steps) and evaluated using 10-fold cross-validation. Briefly, the samples are split into 10 stratified folds (10 disjoint sets in which the proportion of classes in each set approximates, as closely as possible, the proportion in the full dataset). Ten rounds of testing are conducted; for each round, 1 of the 10 folds is held out of the complete dataset, as testing data, and the remaining 9 folds are used to train the model. After model training is complete, the ‘‘held-out’’ testing data are classified, and the numbers of true vs. false positives and negatives are recorded. This process is repeated for all 10 folds of the data, such that every sample is tested (classified) exactly once, and every sample is unknown to the model that classifies it. For gene ontology analysis, we ranked genes by the number of differentially methylated CpGs (P< .05, Wilcoxon’s rank sum test). For grouping genes into sets that were either differentially methylated or not differentially methylated, we first combined individual P values for the CpGs within the promoter region, which we define as 5 kb of the transcription start site, of each RefSeq transcript (53), using the

VOL. - NO. - / - 2015

3

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354

ORIGINAL ARTICLE: ANDROLOGY 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413

method of Stouffer et al. (54). We considered any gene to be differentially methylated if R1 of its transcripts showed differential methylation within this 5-kb window around its transcription start site. Unless otherwise stated, all P values are corrected for multiple hypothesis testing using the method of Benjamini and Hochberg (55).

Global Classifiers

Q10

To capture global properties of DNA methylomes, we constructed models based on the Euclidean distance between samples. We used a simple rule-based classifier (which we call a ‘‘decision stump’’), given as follows: ‘‘If distance to training instance X is greater than Y, classify the sample as class A; else classify it as class B,’’ where A and B depend on the comparison (for example, good vs. poor embryo quality). We define the cost of any such rule as the weighted sum of the number of false positives and false negatives resulting from use of the rule to classify the training data. In the simplest case, each false positive or false negative is given a weight of 1, and the cost is simply the number of samples that are misclassified by the rule on the training data. The values of X and Y are learned by exhaustive enumeration of all possible training instances and discriminative thresholds on the training data; the one with the lowest cost is selected. We modulated the tradeoff between sensitivity and specificity in this learning process by adjusting the relative cost of false positives to false negatives, which we report in the results as the ‘‘false-positive cost’’ of a classification scheme. Schemes that have a high false-positive cost favor results with fewer false positives and hence high specificity. We considered such models to be more clinically valuable than those with high sensitivity.

Site-Specific Classifiers

Q11

Q12

These classifiers were constructed to explore sample classification based on the DNA methylation status (b-values) of a subset of individual CpGs. Here, as before, we used the decision-stump classifier, but in addition, we evaluated a selection of more-complex models and learning algorithms: a support vector machine, a nearest-neighbor classifier, a decision tree, and a naïve Bayes technique. For these models, we used the WEKA implementations (WEKA data mining software), run with default parameters, unless otherwise specified. Further, we augmented several of these with adaptive boosting (AdaBoost; an R software package), and bootstrap aggregating (bagging)—methods designed to avoid overfitting the model to the training data. As with the global classifiers, we favored high-specificity models. For models in which the false-positive cost could not be modulated during training, we weighted the training instances to achieve the same result.

RESULTS Methylation in Purified vs. Unpurified Samples Hierarchic clustering of samples showed that, although purified and unpurified samples were different, purified and unpurified methylomes that were derived from the same sample always clustered with one another; intersample vari-

ations were greater than those among purification methods (Supplemental Fig. 1, available online).

Is Aberrant DNA Methylation Predictive of Fertility Status and Embryo Quality? We first performed hierarchic clustering of all 163 neat samples, based on the methylation level of all of the >485,000 CpGs interrogated by the array. This clustering revealed an ‘‘out-group’’ composed exclusively of IVF patient samples (Fig. 1A, labeled ‘‘patient-only cluster’’), which contained within it a subgroup composed almost exclusively of samples that led to poor embryo quality (labeled ‘‘poor embryo-quality clusters’’ in Fig. 1A). However, most IVF patient samples and poor-quality–embryo samples were distributed evenly among fertile donor samples and good-quality–embryo samples outside this group. To verify the predictive power of the genomewide methylation differences observed from the clustering, we trained a simple decision-stump classifier using 10-fold cross-validation, in which each sample was described by the Euclidean distance to each of the samples in the training set. We adjusted the false-positive cost of the training algorithm (see Materials and Methods section for details) from 1 (i.e., a false positive was considered as costly as a false negative during training) to 10 (i.e., a false positive was considered 10 times more costly than a false negative when training). We observed that with high false-positive costs, the classifier could achieve specificity >0.9, while identifying one third (33%) of poor-quality–embryo samples, and 27% of IVF patient samples (Fig. 1B and C). We replicated this analysis on the density gradient–purified samples and observed an even greater separation of good- vs. poor-quality–embryo samples, although little discernible separation of IVF patient and donor samples was observed, possibly because of the small number of purified donor samples (Fig. 2A). This lack of separation was reflected in the classification results from these samples, which showed even greater classifier identification of poor-quality–embryo samples, with just under one half of the poor-quality–embryo patients recovered, with zero false positives, whereas classifier differentiation between IVF patient and donor samples was largely eliminated (Fig. 2B and C).

Is Predictive Differential Methylation Concentrated Within Specific CpGs or Annotation Categories? To better understand whether differences in methylation between groups were concentrated at particular individual CpGs or regions, we called differentially methylated positions (DMPs) using Wilcoxon’s rank sum test. As a control, we repeated this process using randomly shuffled sample labels. The number of differentially methylated CpGs that were identified between neat IVF patient vs. donor samples, both before and after correction for multiple hypothesis testing, is shown in Figure 3A. The histogram of P values from all CpGs is displayed in Figure 3B; it shows a sharp spike in significant P values for actual labels, with a relatively flat distribution for VOL. - NO. - / - 2015

4

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472

Fertility and Sterility®

FIGURE 1

print & web 4C=FPO

473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531

Aberrant global methylation profiles are indicative of fertility status and poor embryo quality. (A) Hierarchic clustering of 163 neat samples based on a global methylation profile; an ‘‘out-group’’ of exclusively patient samples, with a subgroup strongly enriched for poor embryo samples is apparent. (B) The ROC curve (left) for prediction of patient status (IVF patient or fertile donor). Classification was performed by building a 1-level decision tree (Supplemental Methods) based on Euclidean distance between samples. Training and testing is performed using 10-fold cross-validation. The same information is shown for the task of predicting embryo quality (right). In both cases, high confidence predictions (bottom right of plots) have a high probability of being correct. (C) Classification statistics for the ROC curves are presented in (B). Samples were predicted to be positive (i.e., a patient sample or a poor-quality–embryo sample) when the probability exceeded 50%. In both cases, high specificity and positive predictive value are achieved. AUC ¼ area under the curve; FP ¼ false positive; NPV ¼ negative predictive value; PPV ¼ positive predictive value; Se. ¼ sensitivity; Sp. ¼ specificity. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

the shuffled control. The significant CpGs that were identified using actual labels showed no discernible bias toward particular genomic annotation categories (Fig. 3C). We used the test for differential methylation at individual CpGs to perform feature selection, with the expectation that the CpGs that show the strongest indication of differential methylation would make good features for predictive models. To estimate how reproducible the selection of particular features (CpGs) would be with various datasets, we broke the samples into 10 groups of approximately equal size. For each group, we repeated the selection of DMPs with those samples that had been held out. We observed that the number of groups for which a CpG was selected in the top 100 most differentially methylated CpGs was generally low when using randomly shuffled labels, and follows a Poisson distribution, as expected. In contrast, use of actual labels produced a Ushaped distribution, with a substantially larger proportion of CpGs always appearing in the top 100 (Supplemental Table 3, available online, gives the genes that are intersected by these consistently differentially methylated CpGs). We repeated the classifier training and evaluation described previously, this time using only the top 50,

1,000, 50,000, or 400,000 differentially methylated CpGs in the training data to compute Euclidean distances. Receiver operating characteristic curves for each are displayed in Figure 3E; these show a clear advantage to classification using fewer features. We took this process a step further by training a range of models using the b-values for the selected CpGs as features (rather than Euclidean distance between samples). Using a bagged cost-sensitive support vector machine (see Materials and Methods section), we were able to identify IVF patient samples with a positive predictive value of 99%, and a sensitivity of 82%; receiver operating characteristic curves are shown in Figure 3F. When we applied the same approach to identifying differentially methylated CpGs for good vs. poor-quality—embryo samples (using purified samples, as these showed greater separation in our earlier clustering analysis), we found no significant CpGs after correction for multiple hypothesis testing (Supplemental Fig. 2A, available online). Further, the distribution of P values derived from the actual and shuffled labels did not exhibit the differences observed for IVF patient and donor samples (Supplemental Fig. 2B).

VOL. - NO. - / - 2015

5

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

Q17

532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590

ORIGINAL ARTICLE: ANDROLOGY

FIGURE 2

print & web 4C=FPO

591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649

Gradient purification improves separation of good- and poor-quality sperm samples. (A) Hierarchic clustering of 62 gradient-purified sperm samples based on global methylation profile; 2 clear clusters are apparent, 1 of which contains almost exclusively poor-quality–embryo samples. (B) The ROC curve is shown (left), for prediction of patient status (IVF patient or fertile donor) from the 62 gradient-purified samples. Classification was performed by building a 1-level decision tree (Supplemental Methods) based on Euclidean distance between samples. Training and testing is performed using 10-fold cross-validation. The same information is shown (right) for the task of predicting embryo quality from the subset of 45 samples for which this information is known. Although classification of patient status is degraded from that with unpurified samples, the ability to predict poor-quality embryos is markedly improved. (C) Classification statistics for the ROC curves presented in (B). Samples were predicted as positive (i.e., a patient sample or a poor-quality–embryo sample) when the probability exceeded 50%. In the case of predicting poor-quality embryo, very high specificity and positive predictive value are achieved. AUC ¼ area under the curve; FP ¼ false positive; NPV ¼ negative predictive value; PPV ¼ positive predictive value; Se. ¼ sensitivity; Sp. ¼ specificity. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

As with the DMPs in the neat samples for fertility status, no bias occurred toward regions with any particular annotation category; however, in contrast to the comparison of IVF patients with donors, we saw no tendency for any DMPs to consistently appear in the top 100 when we repeated the feature selection with portions of the data held out, as we did previously (Supplemental Fig. 2C and D). In the training of distance-based classifiers as previously described, we observed that using more CpGs resulted in higher and more-stable specificity and positive predictive value, whereas no discernible pattern in performance was apparent with use of a reduced set of features (Supplemental Fig. 2E). One interpretation of this result is that the poor-quality– embryo samples we are able to differentiate using Euclidean distances create genomewide changes in DNA methylation. However, because we have fewer purified samples, we additionally have reduced statistical power in this setting. Moreover, the platform we have employed here does not give a whole-genome interrogation of methylation. Further studies, including whole-genome bisulfite sequencing, are necessary to more fully explore this finding.

Do Changes in DNA Methylation in CpGs or Promoter Regions Affect Functionally Related Genes? To better understand biological processes and pathways that are affected by the observed changes in DNA methylation, we ranked genes by the number of CpGs that were identified as differentially methylated within their promoter regions (5 kb around the transcription start site) and performed gene ontology analysis on the top 1,000 genes. In all cases (both purified and neat samples for IVF patient vs. fertile donor samples, and good- vs. poor-quality–embryo samples) we observed significant enrichment for genes involved in cell adhesion and morphogenesis (Fig. 4A). Additionally, several gene families associated with embryogenesis and development are differentially methylated in the good vs. poor embryogenesis groups. Following up on this finding, we combined P values within promoter regions for differential methylation between purified good- and poor-quality–embryo samples to generate a single P value for differential methylation of each gene. We VOL. - NO. - / - 2015

6

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

Q18

650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708

Fertility and Sterility®

FIGURE 3

print & web 4C=FPO

709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767

Differential methylation between healthy donor and infertile IVF patient samples consistently occurs at a limited number of CpGs. (A) Number of differentially methylated CpGs for unpurified donor vs. IVF patient samples before and after correction for multiple hypothesis testing, using both the actual donor/patient labels and randomly shuffled donor/patient labels as a control. (B) Distribution of P values for all profiled CpGs from tests of differential methylation between donor and IVF patient samples using both actual labels and randomly shuffled donor/patient labels as a control. (C) Proportion of differentially methylated CpGs (before correction for multiple hypothesis testing) that fall within regions annotated as shown, both actual donor/IVF patient labels and randomly permuted labels. (D) Samples are split into 10 stratified, equal-size groups. A fold is formed by taking 9 of these groups and leaving 1 group out (allowing 10 separate analyses). For each fold, we use the samples in the 9 retained groups to identify differentially methylated CpGs using both actual donor/IVF patient labels and randomly permuted labels as a control. A histogram showing the number of CpGs that were contained in the top 100 most differentially methylated CpGs identified in only 1 fold, exactly 2 folds, exactly 3 folds, and so on, is displayed. The inset shows the number of CpGs contained in all 10 folds, for both actual labels and permuted labels in higher detail. (E) Classifiers, analogous to those in presented in Figure 2 were trained using a subset of the top 50, 1,000, 50,000, and 400,000 most significantly differentially methylated CpGs. The ROC curves are plotted for each. (F) A range of classifiers was trained on the top 500 most differentially methylated CpGs; shown are the ROC curves from these classifiers. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

found that the set of differentially methylated genes identified in this way contained 25 imprinted genes, accounting for close to 10% of known imprinted genes. As a control, we randomly permuted the labels on these samples and repeated the analysis; only 1 imprinted gene was identified as differentially methylated after permutation testing (Fig. 4B).

DISCUSSION The current study evaluated genomewide sperm DNA methylation patterns in a relatively large cohort of normozoospermic, fertile men serving as a control group, and men

undergoing IVF, in an effort to evaluate predictability of fertility status and IVF success based on the sperm DNA methylome. The current study has several important limitations. Replication with a larger control group of less-select fertile men, as well as a larger patient cohort, is required to confirm the findings presented here. In particular, the comparison of methylome signatures using neat samples vs. density gradient–prepared samples, showed that methylation profiles of purified samples were more predictive of embryo quality, whereas those of neat samples were more predictive of fertility status. This finding is intriguing. Purification, in addition to selecting for

VOL. - NO. - / - 2015

7

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826

ORIGINAL ARTICLE: ANDROLOGY

FIGURE 4

print & web 4C=FPO

827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885

Functional classification of differentially methylated genes. (A) Gene ontology analysis of the top 1,000 genes after sorting genes by number of differentially methylated CpGs. All terms that were significant (correct P value <.05), in comparing either IVF patients with fertile donors, or good- and poor-embryogenesis IVF, patient samples are included, with the exception of 24 terms associated with gene expression and cellular metabolism, which are omitted for clarity of visualization (full list is provided in Supplemental Table 3, available online). (B) The absolute number and proportion of genes with differentially methylated promoter regions (P<.01, Wilcoxon’s signed rank test, purified samples) that are known to be imprinted. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

Q13

more-competent sperm, may in addition eliminate immature or incompletely compacted sperm that could indicate poor fertility, possibly accounting for this difference. However, with fewer samples, power in the purified case is reduced, thereby decreasing identification of more-subtle effects, and making direct comparison of results from purified vs. neat samples difficult. Another limitation of this study was the challenge of classifying good vs. poor embryogenesis cases and minimizing inclusion of potentially contributing female infertility factors. Embryogenesis is a complex process, and ‘‘poor embryogenesis’’ can be manifested in several ways, including developmental arrest or attrition at any point during embryonic development, and poor implantation. Moreover, embryogenesis is rarely homogeneous within a cohort of oocytes. Although we made every attempt to select the mostextreme cases (best and worst) while avoiding severe female infertility factors, the cases we selected for the study were heterogeneous by virtue of the inherent complexities of fertilization and embryogenesis. As illustrated in Supplemental Table 1 (available online), although semen parameters and male and female age did not differ between groups, the poor-embryo, negative pregnancy group had, on average, fewer oocytes, metaphase II oocytes, and normally fertilized oocytes. This finding suggests that cryptic overrepresentation of female infertility factors may have occurred in this group and not in the others. Although

these differences are not ideal, they are not expected to affect specificity; assuming sufficient signal for model training, the presence of female factor samples would be expected to result in an increased false-negative rate, thereby reducing sensitivity. Although the predictive power of this approach for classifying fertile and patient samples is remarkably high in this study, the control samples were from carefully screened, normozoospermic, fertile sperm donors, which do not represent the general population, or even the fertile general population. Future studies should include more-representative samples. To our knowledge, this study is the largest on genomewide sperm DNA methylation performed to date. Our findings that sperm DNA methylation patterns are generally very stable across samples from different individuals and across sperm fractions from the same individual are in agreement with those of smaller genomewide sperm DNA methylation studies (44, 45, 47, 48). This study addressed 2 fundamental questions regarding the prognostic value of sperm DNA methylation patterns in the context of male infertility. We first aimed to evaluate the power of sperm DNA methylation patterns, to distinguish normozoospermic, fertile men from infertile men. Second, we investigated the utility of sperm DNA methylation data for predicting IVF outcomes. Our findings suggest that methylation patterns are highly predictive of fertility status, and quite predictive of IVF embryo quality. VOL. - NO. - / - 2015

8

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944

Fertility and Sterility® 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003

Q14

Several relevant, and probably biologically informative, differences were found in the power of methylation data to predict fertility status vs. IVF embryo quality. The primary difference is the methylation signature for each group. Our analyses demonstrated that no individual CpG displayed significantly different methylation in good- vs. poor-quality–embryo groups, after correcting for multiple comparisons. Predictive power was poor when we trained models based on a selected set of features (CpGs), rather than on global profiles. As we increased the number of CpGs for the predictive models, the predictive power increased (Supplemental Fig. 2, available online). This result suggests that although poor embryogenesis cannot be attributed to a few consistently altered CpGs, the genomewide methylation profile seems to be inherently different in men in the poor-quality–embryo group. Another informative result is that with density gradient– purified samples, power to predict good vs. poor embryo quality was significantly improved. Although the reason is unclear, the purification step may have reduced background methylation heterogeneity by eliminating less-competent sperm that, although present in the sample, are unlikely to play a role in fertilization or embryogenesis. This process may thereby have the effect of improving the overall methylation signatures of the 2 groups. The methylation differences in good- vs. poor-embryogenesis samples may be subtle, and dispersed throughout the genome. Additional studies, including whole-genome assays, are necessary to further our understanding of methylation effects. In contrast, comparison of sperm methylomes of fertile men vs. IVF patients revealed, after multiple-comparison correction, >8,500 CpGs that had significantly different methylation. We employed the same strategy for predictive modeling that was used to classify patients with good- vs. poor-quality embryos. In this case, probably driven by the large number of highly differentially methylated CpGs, we found that models were most effective at classifying samples when they were classified using only those CpGs with the most significant methylation differences (Fig. 3). In a comparison of those in the control group vs. IVF patients, whole ejaculate proved much more effective at classifying samples than did the evaluation of methylation in purified samples; however, this difference may be a result of the small number of purified samples in the control group. Evaluation of the genomic context of the methylation alterations, as well as the gene classes affected by differential methylation between groups, indicated no enrichment for differential methylation within a specific genomic context (Figs. 3C and Supplemental Fig. 2). Gene ontology analysis indicated highly significant enrichment for several functional classes of genes (Fig. 5). Most notably, genes involved in cellular adhesion were significantly overrepresented among differentially methylated CpGs, for all comparisons. Although we can only speculate about the relevance of this finding, cell adhesion is known to be critical for both embryogenesis (56, 57) and sperm-oocyte fusion (58, 59). Genes involved in cellular morphogenesis and differentiation were overrepresented in the good vs. poor embryogenesis comparison, and to a lesser extent, in the IVF vs. fertile

donor comparison. We found that imprinted genes were significantly overrepresented among the list of differentially methylated genes in a comparison of good vs. poor embryogenesis samples. This enrichment was not detected in comparison of IVF patients and fertile donors. This is primarily Q15 apparent with purified samples and may explain our observation that methylation status was more strongly related to embryo quality in purified samples; or, these differences may be subtle changes that are masked by other elements in the unpurified samples. In conclusion, multiple studies have identified differences in sperm DNA methylation at single loci, most often imprinted genes, in normozoospermic vs. infertile men, and a few small studies have used array-based approaches to identify sperm DNA methylation differences between the 2 groups. However, this study is the first to exploit the observed differences to build models predictive for fertility status. The findings presented here provide an exciting and potentially clinically useful metric for assessment of male infertility. Additional studies are needed, to replicate the findings presented here, and to evaluate the generalizability of the present findings in a larger cohort of fertile and subfertile populations.

REFERENCES 1.

2. 3.

4.

5.

6. 7.

8.

9. 10.

11.

12.

13.

Q16

Esteves SC, Zini A, Aziz N, Alvarez JG, Sabanegh ES Jr, Agarwal A. Critical appraisal of World Health Organization’s new reference values for human semen characteristics and effect on diagnosis and treatment of subfertile men. Urology 2012;79:16–22. Barratt CL, Mansell S, Beaton C, Tardif S, Oxenham SK. Diagnostic tools in male infertility—the question of sperm dysfunction. Asian J Androl 2011;13:53–8. Bonde JP, Ernst E, Jensen TK, Hjollund NH, Kolstad H, Henriksen TB, et al. Relation between semen quality and fertility: a population-based study of 430 first-pregnancy planners. Lancet 1998;352:1172–7. Guzick DS, Overstreet JW, Factor-Litvak P, Brazil CK, Nakajima ST, Coutifaris C, et al. Sperm morphology, motility, and concentration in fertile and infertile men. N Engl J Med 2001;345:1388–93. Gandini L, Menditto A, Chiodo F, Lenzi A. From the European Academy of Andrology. Italian pilot study for an external quality control scheme in semen analysis and antisperm antibiotics detection. Int J Androl 2000;23:1–3. Neuwinger J, Behre HM, Nieschlag E. External quality control in the andrology laboratory: an experimental multicenter trial. Fertil Steril 1990;54:308–14. Brazil C, Swan SH, Tollner CR, Treece C, Drobnis EZ, Wang C, et al. Quality control of laboratory methods for semen evaluation in a multicenter research study. J Androl 2004;25:645–56. Leushuis E, van der Steeg JW, Steures P, Repping S, Bossuyt PM, Blankenstein MA, et al. Reproducibility and reliability of repeated semen analyses in male partners of subfertile couples. Fertil Steril 2010;94:2631–5. World Health Organization (WHO). WHO laboratory manual for the examination and processing of human semen. 5th ed. Geneva: WHO; 2010. Cooper TG, Noonan E, von Eckardstein S, Auger J, Baker HW, Behre HM, et al. World Health Organization reference values for human semen characteristics. Hum Reprod Update 2010;16:231–45. Zini A, Boman JM, Belzile E, Ciampi A. Sperm DNA damage is associated with an increased risk of pregnancy loss after IVF and ICSI: systematic review and meta-analysis. Hum Reprod 2008;23:2663–8. Sanchez R, Toepfer-Petersen E, Aitken RJ, Schill WB. A new method for evaluation of the acrosome reaction in viable human spermatozoa. Andrologia 1991;23:197–203. Lee MA, Trucco GS, Bechtol KB, Wummer N, Kopf GS, Blasco L, et al. Capacitation and acrosome reactions in human spermatozoa monitored by a chlortetracycline fluorescence assay. Fertil Steril 1987;48:649–58.

VOL. - NO. - / - 2015

9

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062

ORIGINAL ARTICLE: ANDROLOGY 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121

14. 15.

16.

17.

18.

19. 20.

21. 22.

23.

24.

25.

26.

27. 28. 29.

30. 31.

32.

33.

34.

35.

36.

Haas GG Jr, Cines DB, Schreiber AD. Immunologic infertility: identification of patients with antisperm antibody. N Engl J Med 1980;303:722–7. Burkman LJ, Coddington CC, Franken DR, Krugen TF, Rosenwaks Z, Hogen GD. The hemizona assay (HZA): development of a diagnostic test for the binding of human spermatozoa to the human hemizona pellucida to predict fertilization potential. Fertil Steril 1988;49:688–97. Yanagimachi R, Yanagimachi H, Rogers BJ. The use of zona-free animal ova as a test-system for the assessment of the fertilizing capacity of human spermatozoa. Biol Reprod 1976;15:471–6. Bartoov B, Berkovitz A, Eltes F, Kogosowski A, Menezo Y, Barak Y. Real-time fine morphology of motile human sperm cells is associated with IVF-ICSI outcome. J Androl 2002;23:1–8. Huszar G, Ozenci CC, Cayli S, Zavaczki Z, Hansch E, Vigue L. Hyaluronic acid binding by human sperm indicates cellular maturity, viability, and unreacted acrosomal status. Fertil Steril 2003;79(Suppl 3):1616–24. Sigman M, Baazeem A, Zini A. Semen analysis and sperm function assays: What do they mean? Semin Reprod Med 2009;27:115–23. Tiepolo L, Zuffardi O. Localization of factors controlling spermatogenesis in the nonfluorescent portion of the human Y chromosome long arm. Hum Genet 1976;34:119–24. Lanfranco F, Kamischke A, Zitzmann M, Nieschlag E. Klinefelter’s syndrome. Lancet 2004;364:273–83. Dam AH, Koscinski I, Kremer JA, Moutou C, Jaeger AS, Oudakker AR, et al. Homozygous mutation in SPATA16 is associated with male infertility in human globozoospermia. Am J Hum Genet 2007;81:813–20. Elinati E, Kuentz P, Redin C, Jaber S, Vanden Meerschaut F, Makarian J, et al. Globozoospermia is mainly due to DPY19L2 deletion via non-allelic homologous recombination involving two recombination hotspots. Hum Mol Genet 2012;21:3695–702. Harbuz R, Zouari R, Pierre V, Ben Khelifa M, Kharouf M, Coutton C, et al. A recurrent deletion of DPY19L2 causes infertility in man by blocking sperm head elongation and acrosome formation. Am J Hum Genet 2011;88:351–61. Koscinski I, Elinati E, Fossard C, Redin C, Muller J, Velez de la Calle J, et al. DPY19L2 deletion as a major cause of globozoospermia. Am J Hum Genet 2011;88:344–50. Aoki VW, Liu L, Jones KP, Hatasaka HH, Gibson M, Peterson CM, et al. Sperm protamine 1/protamine 2 ratios are related to in vitro fertilization pregnancy rates and predictive of fertilization ability. Fertil Steril 2006;86: 1408–15. Aoki VW, Moskovtsev SI, Willis J, Liu L, Mullen JB, Carrell DT. DNA integrity is compromised in protamine-deficient human sperm. J Androl 2005;26:741–8. Balhorn R, Reed S, Tanphaichitr N. Aberrant protamine 1/protamine 2 ratios in sperm of infertile human males. Experientia 1988;44:52–5. Carrell DT, Liu L. Altered protamine 2 expression is uncommon in donors of known fertility, but common among men with poor fertilizing capacity, and may reflect other abnormalities of spermiogenesis. J Androl 2001;22:604–10. Perreault SD. Chromatin remodeling in mammalian zygotes. Mutat Res 1992;296:43–55. Hammoud SS, Nix DA, Zhang H, Purwar J, Carrell DT, Cairns BR. Distinctive chromatin in human sperm packages genes for embryo development. Nature 2009;460:473–8. Miller D, Brinkworth M, Iles D. Paternal DNA packaging in spermatozoa: more than the sum of its parts? DNA, histones, protamines and epigenetics. Reproduction 2010;139:287–301. Benchaib M, Ajina M, Lornage J, Niveleau A, Durand P, Guerin JF. Quantitation by image analysis of global DNA methylation in human spermatozoa and its prognostic value in in vitro fertilization: a preliminary study. Fertil Steril 2003;80:947–53. Houshdaran S, Cortessis VK, Siegmund K, Yang A, Laird PW, Sokol RZ. Widespread epigenetic abnormalities suggest a broad DNA methylation erasure defect in abnormal human sperm. PLoS One 2007;2:e1289. Kobayashi H, Sato A, Otsu E, Hiura H, Tomatsu C, Utsunomiya T, et al. Aberrant DNA methylation of imprinted loci in sperm from oligospermic patients. Hum Mol Genet 2007;16:2542–51. Marques CJ, Costa P, Vaz B, Carvalho F, Fernandes S, Barros A, et al. Abnormal methylation of imprinted genes in human sperm is associated with oligozoospermia. Mol Hum Reprod 2008;14:67–74.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48. 49.

50.

51.

52.

53.

54. 55. 56. 57. 58. 59.

Kobayashi H, Hiura H, John RM, Sato A, Otsu E, Kobayashi N, et al. DNA methylation errors at imprinted loci after assisted conception originate in the parental sperm. Eur J Hum Genet 2009;17:1582–91. Chen SL, Shi XY, Zheng HY, Wu FR, Luo C. Aberrant DNA methylation of imprinted H19 gene in human preimplantation embryos. Fertil Steril 2010;94: 2356–8. 2358.e1. Hammoud SS, Purwar J, Pflueger C, Cairns BR, Carrell DT. Alterations in sperm DNA methylation patterns at imprinted loci in two classes of infertility. Fertil Steril 2010;94:1728–33. Navarro-Costa P, Nogueira P, Carvalho M, Leal F, Cordeiro I, Calhaz-Jorge C, et al. Incorrect DNA methylation of the DAZL promoter CpG island associates with defective human sperm. Hum Reprod 2010;25:2647–54. Poplinski A, Tuttelmann F, Kanber D, Horsthemke B, Gromoll J. Idiopathic male infertility is strongly associated with aberrant methylation of MEST and IGF2/H19 ICR1. Int J Androl 2010;33:642–9. Sato A, Hiura H, Okae H, Miyauchi N, Abe Y, Utsunomiya T, et al. Assessing loss of imprint methylation in sperm from subfertile men using novel methylation polymerase chain reaction Luminex analysis. Fertil Steril 2011;95:129–34. Nanassy L, Carrell DT. Abnormal methylation of the promoter of CREM is broadly associated with male factor infertility and poor sperm quality but is improved in sperm selected by density gradient centrifugation. Fertil Steril 2011;95:2310–4. Pacheco SE, Houseman EA, Christensen BC, Marsit CJ, Kelsey KT, Sigman M, et al. Integrative DNA methylation and gene expression analyses identify DNA packaging and epigenetic regulatory genes associated with low motility sperm. PLoS One 2011;6:e20280. Aston KI, Punj V, Liu L, Carrell DT. Genome-wide sperm deoxyribonucleic acid methylation is altered in some men with abnormal chromatin packaging or poor in vitro fertilization embryogenesis. Fertil Steril 2012;97:285–92.e4. Klaver R, Tuttelmann F, Bleiziffer A, Haaf T, Kliesch S, Gromoll J. DNA methylation in spermatozoa as a prospective marker in andrology. Andrology 2013;1:731–40. Krausz C, Sandoval J, Sayols S, Chianese C, Giachini C, Heyn H, et al. Novel insights into DNA methylation features in spermatozoa: stability and peculiarities. PloS One 2012;7:e44479. Jenkins TG, Aston KI, Trost C, Farley J, Hotaling JM, Carrell DT. Intra-sample heterogeneity of sperm DNA methylation. Mol Hum Reprod 2015;21:313–9. Jenkins TG, Aston KI, Pflueger C, Cairns BR, Carrell DT. Age-associated sperm DNA methylation alterations: possible implications in offspring disease susceptibility. PLos Genet 2014;10:e1004458. Nanassy L, Peterson CA, Wilcox AL, Peterson CM, Hammoud A, Carrell DT. Comparison of 5% and ambient oxygen during days 3-5 of in vitro culture of human embryos. Fertil Steril 2010;93:579–85. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 2014;30:1363–9. Maksimovic J, Gordon L, Oshlack A. SWAN: subset-quantile within array normalization for illumina infinium HumanMethylation450 BeadChips. Genome Biol 2012;13:R44. Pruitt KD, Brown GR, Hiatt SM, Thibaud-Nissen F, Astashyn A, Ermolaeva O, et al. RefSeq: an update on mammalian reference sequences. Nucleic Acids Res 2014;42:D756–63. Stouffer SA, Suchman EA, DeVinney LC, Star SA, Williams RM. Adjustment during Army life. Princeton, NJ: Princeton University Press; 1949. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statis Soc, Ser B 1995;57:289–300. Thiery JP, Duband JL, Rutishauser U, Edelman GM. Cell adhesion molecules in early chicken embryogenesis. Proc Natl Acad Sci U S A 1982;79:6737–41. Ekblom P, Vestweber D, Kemler R. Cell-matrix interactions and cell adhesion during development. Annu Rev Cell Biol 1986;2:27–47. Wassarman PM. Mammalian fertilization: molecular aspects of gamete adhesion, exocytosis, and fusion. Cell 1999;96:175–83. Talbot P, Shur BD, Myles DG. Cell adhesion and fertilization: steps in oocyte transport, sperm-zona pellucida interactions, and sperm-egg fusion. Biol Reprod 2003;68:1–9.

VOL. - NO. - / - 2015

10

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:16 pm
ce BAS

1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180

Fertility and Sterility®

SUPPLEMENTAL FIGURE 1

web 4C=FPO

1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239

Hierarchic clustering of methylation profiles for the 44 IVF patients for which both purified and unpurified samples were processed (i.e., 88 samples are plotted). For each unpurified sample, its nearest neighbor is the purified sample from the same patient. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

VOL. - NO. - / - 2015

10.e1

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:17 pm
ce BAS

Q19

1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298

ORIGINAL ARTICLE: ANDROLOGY

SUPPLEMENTAL FIGURE 2

web 4C=FPO

1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357

Differential methylation between purified good and poor embryogenesis samples is not defined by a small group of consistent single-CpG differences. (A) Number of differentially methylated CpGs of purified good vs. poor embryogenesis samples before and after correction for multiple hypothesis testing, using both the actual good/poor labels and randomly shuffled good/poor labels as a control. (B) Distribution of P values for all profiled CpGs from test of differential methylation between good and poor embryogenesis, using both actual labels and randomly shuffled good/poor labels as a control. (C) Proportion of differentially methylated CpGs (before correction for multiple hypothesis testing) that fall within regions annotated as shown, both actual good/poor embryo quality labels and randomly permuted labels. (D) Samples are split into 10 stratified, equal-size groups. A fold is formed by taking 9 of these groups and leaving 1 group out (allowing 10 ways of doing this). For each fold, we use the samples in the 9 retained groups to identify differentially methylated CpGs, using both actual good/poor embryogenesis labels and randomly permuted labels as a control. A histogram showing the number of CpGs that were contained in the top 100 most differentially methylated CpGs identified in only 1 fold, exactly 2 folds, exactly 3 folds, and so on, is shown. (E) Classifiers, analogous to those in presented in Figure 2, were trained using a subset of the top x most differentially methylated CpGs (where x is varied along the x-axis of the plots), and the sensitivity, specificity, positive predictive value, and negative predictive value of each was evaluated as a function of how many CpGs were selected. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

VOL. - NO. - / - 2015

10.e2

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:17 pm
ce BAS

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416

Fertility and Sterility® 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475

SUPPLEMENTAL TABLE 1 General composition of the study groups, including semen parameters and IVF embryo quality. Good embryos, pregnant Poor embryos, not pregnant Poor embryos, pregnant P value Male age (y) Sperm concentration (M/ml) Progressively motile sperm (%) Female age (y) Eggs retrieved, n MII eggs, n Eggs fertilized normally, n Embryos cryopreserved, n Fertilized eggs Rlevel 2: 6-cell on day 3 (%) Fertilized eggs Rlevel 2: early blast on day 5/6 (%) Embryos transferred, n

33.08 58.58 45.00 30.93 14.07 12.09 10.74 4.21 0.73 0.44 2.04

5.97 62.78 20.15 4.52 5.91 5.15 4.82 3.10 0.23 0.18 0.55

32.98 69.31 42.64 32.60 10.47 8.87 7.23 0.26 0.37 0.07 2.17

4.29 77.13 18.29 5.12 4.13 3.89 3.33 1.16 0.30 0.09 0.80

33.15 76.26 43.44 30.98 12.98 11.29 9.26 0.81 0.45 0.13 2.34

5.03 71.60 24.86 4.51 4.72 3.83 4.00 1.55 0.30 0.15 0.53

.9913 .4527 .8867 .242 .0103 .0073 .0019 < .0001 < .0001 < .0001 .0596

Note: P values are from ANOVA. Values for each embryo/pregnancy category are mean followed by SD. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

VOL. - NO. - / - 2015

10.e3

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:17 pm
ce BAS

1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534

ORIGINAL ARTICLE: ANDROLOGY 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593

SUPPLEMENTAL TABLE 2 Frequency of male and mild female factors in 127 IVF patient-couples. Factor

Good embryogenesis, positive pregnancy (n [ 53)

Poor embryogenesis, negative pregnancy (n [ 31)

Poor embryogenesis, positive pregnancy (n [ 42)

15 16 16 6 10 2 8

12 13 3 3 3 5 3

14 17 6 5 10 3 3

Male factor only Female factor only Male and female factor Idiopathic (unexplained) Endometriosis Diminished ovarian reserve Polycystic ovary syndrome Note: No significant differences were observed among groups. Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

VOL. - NO. - / - 2015

10.e4

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:17 pm
ce BAS

1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652

Fertility and Sterility® 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711

1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770

SUPPLEMENTAL TABLE 3 Genes that overlap CpGs consistently in the top 100 most differentially methylated CpGs, in comparison of unpurified donor vs. IVF patient samples. Gene AHDC1 ALOX5AP BTBD17 CXXC11 EEF1A2 FBLN2 FGF18 GRM6 HIST1H4J HIST1H4K INPP5A JAG2 KCNQ1 KIAA0319L LRRC45 MIR4734 MLLT6 MTMR6 MXRA7 NCDN NDUFS6 NDUFS8 OGFOD2 PSTPIP1 RAP1GAP2 SERPINF2 STRA13 SYT8 TCIRG1 TNNI2 USP24 Aston. Sperm DNA methylation and infertility. Fertil Steril 2015.

VOL. - NO. - / - 2015

10.e5

FLA 5.4.0 DTD
FNS29847_proof
7 September 2015
5:17 pm
ce BAS